Comparison of dominant contributions calculated in Gaussian and Turbomole

[doctor]

Recently, I began use Turbomole for vertical excitation calculations. Upon comparison of obtained results with the data obtained from Gaussian 03 calculations, I find out that energies and transitions oscillator strengths are matched very well, but dominant contributions are not matched, especially in percentage ratio.
As an example, I show the results for the 4th singlet excitation of nitrobenzene.
Gaussian 03 (TD-B3LYP//cc-pVTZ):

Excited State  4:   Singlet-A'     4.8623 eV  254.99 nm  f=0.1961
      31 -> 33         0.64665
      32 -> 34        -0.12661


Turbomole 5.10 (TD-b3lyp_Gaussian//cc-pVTZ):

4 singlet a excitation

Excitation energy / ev:  4.8637
Oscillator strength:
Mixed representation: 0.1968

Dominant contributions:
Occ. orbital   energy/ev   virt.orbital   energy/ev   |coeff.| ^2*100
31a   -7.98   33 a   -2.67   95.2


Do you have any ideas why this might be the case?

[antti_karttunen]

Hello,

One reason to the discrepancy might be that the contributions printed out by Gaussian cannot be directly compared with those printed out with Turbomole

I'll cite a very recent paper entitled "Spectroscopic properties of cyclometallated iridium complexes by TDDFT" (http://dx.doi.org/10.1016/j.theochem.2009.07.025):

"For closed shell molecules, the Gaussian 03 TDDFT vectors are actually normalized to 1/2, so we take the double of the squared coefficients"

This procedure is also discussed in the following CCL post:  http://www.ccl.net/chemistry/resources/messages/2006/07/31.002-dir/index.html

So, the contribution

31 -> 33         0.64665

in terms of percentages would be

0.64665^2 * 2 * 100 = 83.6%

Now, I'd say that this is already much better in comparison to the 95.2 obtained in the case of Turbomole. I'm not sure how easily a 100% agreement can be reached, or if it is possible at all. There are many other things that could result in differences (DFT grid, etc.), but then again, the excitation energies do appear to be almost identical.

Regards,
Antti